Elasticsearch : Hight disk read + slow indexing

I am doing a benchmark to index 200GB of simple structered document (without text field, without nested document).
I Observe periodic degradation on indexing rate during at least one hour and hight disk read io. I use bulk request with refresh_interval = -1, Monitoring not show cpu or memory problem.
I suspect flush or merge process but with monitoring (_stats/merge,flush) i do not see increase of merge or flush.

I use one node with 5 shard, Memory 8GB and 4 cpu (disk hdd).

Here are the monitoring diagrams :

indexing_time operation_rate request_rate segment_count

How to monitore which process of elasticsearch do massive disk read (80mb/s during long period) ?
How to improve indexing rate and keep constant indexing rate with elasticsearch?


This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.